Patent Grant
7085323
All of the material in this patent application is subject to copyright protection under the copyright laws of the United States and of other countries. As of the first effective filing date of the present application, this material is protected as unpublished material. However, permission to copy this material is hereby granted to the extent that the copyright owner has no objection to the facsimile reproduction by anyone of the patent documentation or patent disclosure, as it appears in the United States Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.
This invention relates to the field of digital enhancement of video images.
Enhancements in display resolution has made it possible for many viewers to see video having resolutions greater than the initial recorded or received video images. For example, many television screens are capable of greater pixel resolution than the pixel resolution transmitted with standard broadcast programming or recorded on videotape or disc. Furthermore, many MPEG videos have display resolutions substantially lower than the display resolution of monitors presenting the video.
The advent of modern digital signal processing circuits and algorithms have resulted in processes for enhancing resolution of still images by processing a series of initial resolution images and determining differences there between. These differences are used to build enhanced resolution images having greater pixel density. The term “superresolution” (SR) image enhancement is used by those familiar with this area of the art to describe such image enhancement.
In this example, high resolution image Hk 260 is reconstructed by using five low resolution frames
Lt,tετk∪{k} where
τk={k−2, k−1, k+1, k+2} is the frame index set of the temporal related frames.
While the SR frame processor does produce high resolution still frames, the process has a high computational load and is not optimally suited for generation of high resolution video sequences.
Thus, what is needed is a method and apparatus for constructing an enhanced resolution video image that efficiently builds upon SR techniques for generation of high resolution still frames. The construction should further take advantage of the high resolution video sequence itself in order to reduce the computational loading of its construction.
The invention includes a method of producing an enhanced video image from an initial video image. The method includes the step enhancing resolution of an initial image component common to a plurality of temporal frames of the initial video image by processing the plurality of frames of the received video image to produce an enhanced image component. The method further includes the step of incorporating said enhanced image component into a plurality of temporal frames of said enhanced video image.
The invention includes a device for producing an enhanced video image from an initial video image. The device comprises a key frame generator for enhancing resolution of an initial image component common to a plurality of initial temporal frames of the initial video image by processing the plurality of initial temporal frames to produce an enhanced image component and for incorporating the enhanced image component into a key frame included within a plurality of enhanced temporal frames of the enhanced video image, the key frame corresponding to one of the plurality of initial temporal frames. The device also comprises a non-key frame generator coupled to said key frame generator for incorporating the enhanced image component into a non-key frame included within the plurality of enhanced temporal frames, the non-key frame corresponding to another of the plurality of initial temporal frames.
The invention includes a computer readable medium including computer instructions for producing an enhanced video image from an initial video image. The computer instructions include instructions for enhancing resolution of an initial image component common to a plurality of temporal frames of the initial video image by processing the plurality of frames of the received video image to produce an enhanced image component. The computer instructions also include instructions for incorporating said enhanced image component into a plurality of temporal frames of said enhanced video image.
It is important to note, that these embodiments are only examples of the many advantageous uses of the innovative teachings herein. In general, statements made in the specification of the present application do not necessarily limit any of the equivalent and various claimed inventions. Moreover, some statements may apply to some inventive features but not to others. In general, unless otherwise indicated, singular elements may be in the plural and visa versa with no loss of generality.
The invention reduces the computational loading of high resolution video construction by construction fewer “key frames” using the superresolution frame processor and synthesizing non-key frames using techniques requiring less computational loading. Non-key frames incorporate high resolution image components generated in key frames. Thus, a high resolution image component is incorporated into a plurality of temporal frames such as the high resolution key frame and at least one high resolution non-key frame.
Assume the high resolution key frame Hk has already been constructed by applying a SR still technique on low resolution frame Lk and its temporal related low resolution frames Lt(tετk). A process to synthesizing the high resolution non-key frame Ht(tετk) through image warping techniques is discussed. In an ideal video case, the high resolution image Ht can be obtained by applying the true subpixel motion information from Lt to Lk, M(t→k) on the constructed high resolution image Hk.
Ht=M(t→k)(Hk),∀ετk
However, the true motion information is not available. Therefore the SR image Ht can be approximated by applying the estimated subpixel motion information from Lt to Lk ({circumflex over (M)}(t→k)) on the high resolution image Hk.
Ĥt={circumflex over (M)}(t→k)(Hk),∀tετk
where Ĥt corresponds to image warping and denotes the approximated version of high resolution image Ht.
The error of this approximation is:
E1=Ht−Ĥt,∀tετk
y=M(t→k)(x)
ŷ={circumflex over (M)}(t→k)(x)
Therefore, the true intensity value at pixel x in image Ht is:
Ht(x)=Hk(y)
The estimated intensity value at pixel x in image Ht is:
Ĥt(x)=Hk(ŷ)
which results in synthesis of an image component x of a non-key frame using image warping. The error approximation at pixel x is:
E1(x)=Hk(y)−Hk(ŷ).
If the estimated motion information is accurate enough, the above approach of Ĥt(x)=Hk(ŷ) gives very good results with dramatically decreased computational load compared to the ‘sliding window’ approach of
If motion error is out of range the image component from It is substituted for the image component from Hk. That is:
Ĥt(x)=It(x).
Thus, if the motion error component is determined within range, then warping of the key frame image component can be used where Ĥt(x)=Hk(ŷ), ŷ={circumflex over (M)}(t→k)(x) However, if the motion error component is out of range, then substitution of the blurred non-key frame image can be used where Ĥt(x)=It(x).
If the motion error is not out of range then blending can be used. Blending is an enhancement to warping and blends the information from the initial upsampled image component It with the corresponding image component from the high resolution key frame Hk. That is:
Ĥt(x)=ω1It(x)+ω2Hk(ŷ)
As previously mentioned ω1+ω2=1, and the more accurate the motion information the more predominant the image component from the high resolution key frame Hk. However, as the motion information becomes less accurate the lower resolution upsampled image component from It becomes more predominant.
The accuracy of the motion information is judged by mean square error (MSE) or sum of absolute difference (SAD). Let B1 be a small block N×N (N=5, or 7) centered at pixel x in frame It, let B2 be a small block (same size as B1) centered at pixel ŷ in frame Ik,
The smaller MSE or SAD, the more accurate motion estimation at pixel x; and vice versa. Weight coefficients ω1 and ω2 can be computed as follows:
ω1=1−ω2
Here T1 or T2 is a small threshold, when MSE<T1 or SAD<T2, the motion estimation is assumed accurate.
The warped image component Ĥt(x) may be further improved with a weighted error correction factor ΔHt(x). Error correction is preferably used where ω1≈0. Error correction results in an error corrected image component {tilde over (H)}t(x) where
{tilde over (H)}t(x)=Ĥt(x)+ΔHt(x) and
ΔHt(x)=ω(It(x)−Ik(ŷ)).
The error correction arise from the presumption that the true motion of an image component from Ht to Hk is substantially the same as the motion of the image component from It to Ik. ΔHt(x) provides for pixel or image component intensity adjustments. The value ω is a weighing factor for determining the amount of error correction applied. In order to determine ω for a minimal error an error term is defined.
E2(x)=Ht(x)−{tilde over (H)}t(x)=(Hk(y)−Hk(ŷ))−ω(Ik(y)−Ik(ŷ))
Ideally if E2=0, then
However, the true mapping of image component y in Hk for image component x in Ht is not available, thus two alternative ways to reduce E2 are described.
In the first alternative, let ω=1. This arises from the assumption that the changes of intensity value at image component y and ŷ between the initial upsampled image Ik and the high resolution image Hk are similar. That is:
Hk(y)≈Ik(y)+α and
Hk(ŷ)≈Ik(ŷ)+α thus
ω≈1.
This results in an error corrected and blended image component of:
{tilde over (H)}t(x)=Ĥt(x)+It(x)−Ik(ŷ).
In the second alternative, let
This arises from the assumption that changes of the intensity value an image components y and ŷ between the initial upsampled image Ik and the high resolution image Hk can be approximated by scalar c. That is:
Hk(y)≈cIk(y) and
Hk(ŷ)≈cIk(ŷ) thus
This results in an error corrected and blended image component of:
It should be further noted that c can alternately be chosen to extend
to an image component that includes a small neighborhood around ŷ.
Thus, what is provided is a method and apparatus for constructing an enhanced resolution video image that efficiently builds upon SR techniques for generation of high resolution still frames. The construction further takes advantage of the high resolution video frame sequence itself in order to reduce the computational complexity of its construction. Key frames are generated using SR techniques. Non-key frames are generated using image warping. Image warping further improved using blending and error correction techniques.
According to the inventive principles as disclosed in connection with the preferred embodiment, the invention and the inventive principles are not limited to any particular kind of computer system but may be used with any general purpose computer, as would be known to one familiar with the art, arranged to perform the functions described and the method steps described. The operations of such a computer, as described above, may be according to a computer program contained on a medium for use in the operation or control of the computer, as would be known to one familiar with the art. The computer medium that may be used to hold or contain the computer program product, may be a fixture of the computer such as an embedded memory or may be on a transportable medium such as a disk, as would be known to one familiar with the art.
The invention is not limited to any particular computer program or logic or language, or instruction but may be practiced with any such suitable program, logic or language, or instructions as would be known to one familiar with the art. Without limiting the principles of the disclosed invention any such computing system can include, inter alia, at least a computer readable medium allowing a computer to read data, instructions, messages or message packets, and other computer readable information from the computer readable medium. The computer readable medium may include non-volatile memory, such as ROM, Flash memory, floppy disk, Disk drive memory, CD-ROM, and other permanent storage. Additionally, a computer readable medium may include, for example, volatile storage such as RAM, buffers, cache memory, and network circuits.
Furthermore, the computer readable medium may include computer readable information in a transitory state medium such as a network link and/or a network interface, including a wired network or a wireless network, that allow a computer to read such computer readable information.
Furthermore, the invention may be embodied in an electrical circuit receiving low resolution signals and producing high resolution signals. As an example, circuit may be included within as a set-top-box interposed between a television monitor display for displaying the superresolution image and a signal source such as an antenna, cable TV system, VCR or satellite receiver. The set top box is detachable and removable because the it may be decoupled from the display and signal source, thereby facilitating display of non-superresolution video. Alternately the electrical circuit may be included within a printed circuit board. The printed circuit board may be installed within a television, personal computer having a standard connector for accessory cards or other image display system. In another example the electrical circuit may be included in one or more integrated circuits such as application specific integrated circuits, custom integrated circuits or arrangements of off-the-shelf integrated circuits (including microprocessors) or combinations thereof. Furthermore, the invention may be incorporation into a combination of one or more electrical circuits and computer readable information, instructions or software.
Although a specific embodiment of the invention has been disclosed. It will be understood by those familiar with the art that changes can be made to this specific embodiment without departing from the spirit and scope of the invention. The scope of the invention is not to be restricted, therefore, to the specific embodiment, and it is intended that the appended claims cover any and all such applications, modifications, and embodiments within the scope of the present invention. For example, an image component is preferably associated with a high resolution pixel, but could alternately refer to a grouping of high resolution pixels or one or more sub-pixels of a higher resolution image portion.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5579054 | Sezan et al. | Nov 1996 | A |
| 5696848 | Patti et al. | Dec 1997 | A |
| 5808695 | Rosser et al. | Sep 1998 | A |
| 5969772 | Saeki | Oct 1999 | A |
| 6023535 | Aoki | Feb 2000 | A |
| 6069668 | Woodham, Jr. et al. | May 2000 | A |
| 6181826 | Weldy et al. | Jan 2001 | B1 |
| 6208765 | Bergen | Mar 2001 | B1 |
| 6259828 | Crinon et al. | Jul 2001 | B1 |
| 6304682 | Patti | Oct 2001 | B1 |
| 6349154 | Kleihorst | Feb 2002 | B1 |
| 6393163 | Burt et al. | May 2002 | B1 |
| 6782132 | Fogg | Aug 2004 | B1 |
| Number | Date | Country | |
|---|---|---|---|
| 20030189983 A1 | Oct 2003 | US |
